Abstract: Cold-start is a notoriously difficult problem which
can occur in recommendation systems, and arises when there is
insufficient information to draw inferences for users or items. To
address this challenge, a contextual bandit algorithm – the Fast
Approximate Bayesian Contextual Cold Start Learning algorithm
(FAB-COST) – is proposed, which is designed to provide improved
accuracy compared to the traditionally used Laplace approximation
in the logistic contextual bandit, while controlling both algorithmic
complexity and computational cost. To this end, FAB-COST uses
a combination of two moment projection variational methods:
Expectation Propagation (EP), which performs well at the cold
start, but becomes slow as the amount of data increases; and
Assumed Density Filtering (ADF), which has slower growth of
computational cost with data size but requires more data to obtain an
acceptable level of accuracy. By switching from EP to ADF when
the dataset becomes large, it is able to exploit their complementary
strengths. The empirical justification for FAB-COST is presented, and
systematically compared to other approaches on simulated data. In a
benchmark against the Laplace approximation on real data consisting
of over 670, 000 impressions from autotrader.co.uk, FAB-COST
demonstrates at one point increase of over 16% in user clicks. On
the basis of these results, it is argued that FAB-COST is likely to
be an attractive approach to cold-start recommendation systems in a
variety of contexts.
Abstract: In recent years, e-learning recommender systems has attracted great attention as a solution towards addressing the problem of information overload in e-learning environments and providing relevant recommendations to online learners. E-learning recommenders continue to play an increasing educational role in aiding learners to find appropriate learning materials to support the achievement of their learning goals. Although general recommender systems have recorded significant success in solving the problem of information overload in e-commerce domains and providing accurate recommendations, e-learning recommender systems on the other hand still face some issues arising from differences in learner characteristics such as learning style, skill level and study level. Conventional recommendation techniques such as collaborative filtering and content-based deal with only two types of entities namely users and items with their ratings. These conventional recommender systems do not take into account the learner characteristics in their recommendation process. Therefore, conventional recommendation techniques cannot make accurate and personalized recommendations in e-learning environment. In this paper, we propose a recommendation technique combining collaborative filtering and ontology to recommend personalized learning materials to online learners. Ontology is used to incorporate the learner characteristics into the recommendation process alongside the ratings while collaborate filtering predicts ratings and generate recommendations. Furthermore, ontological knowledge is used by the recommender system at the initial stages in the absence of ratings to alleviate the cold-start problem. Evaluation results show that our proposed recommendation technique outperforms collaborative filtering on its own in terms of personalization and recommendation accuracy.
Abstract: Transient simulation of the hydrogen-assisted self-ignition of propane-air mixtures were carried out in platinum-coated micro-channels from ambient cold-start conditions, using a two-dimensional model with reduced-order reaction schemes, heat conduction in the solid walls, convection and surface radiation heat transfer. The self-ignition behavior of hydrogen-propane mixed fuel is analyzed and compared with the heated feed case. Simulations indicate that hydrogen can successfully cause self-ignition of propane-air mixtures in catalytic micro-channels with a 0.2 mm gap size, eliminating the need for startup devices. The minimum hydrogen composition for propane self-ignition is found to be in the range of 0.8-2.8% (on a molar basis), and increases with increasing wall thermal conductivity, and decreasing inlet velocity or propane composition. Higher propane-air ratio results in earlier ignition. The ignition characteristics of hydrogen-assisted propane qualitatively resemble the selectively inlet feed preheating mode. Transient response of the mixed hydrogen- propane fuel reveals sequential ignition of propane followed by hydrogen. Front-end propane ignition is observed in all cases. Low wall thermal conductivities cause earlier ignition of the mixed hydrogen-propane fuel, subsequently resulting in low exit temperatures. The transient-state behavior of this micro-scale system is described, and the startup time and minimization of hydrogen usage are discussed.
Abstract: Presently, engine cooling pump is driven by toothed
belt. Therefore, the pump speed is dependent on engine speed which
varies their output. At normal engine operating conditions (Higher
RPM and low load, Higher RPM and high load), mechanical water
pumps in existing engines are inevitably oversized and so the use of
an electric water pump together with state-of-the-art thermal
management of the combustion engine has measurable advantages.
Demand-driven cooling, particularly in the cold-start phase, saves
fuel (approx 3 percent) and leads to a corresponding reduction in
emissions. The lack of dependence on a mechanical drive also results
in considerable flexibility in component packaging within the engine
compartment. This paper describes the testing and comparison of
existing mechanical water pump with that of the electric water pump.
When the existing mechanical water pump is replaced with the new
electric water pump the percentage gain in system efficiency is also
discussed.
Abstract: We propose an enhanced collaborative filtering
method using Hofstede-s cultural dimensions, calculated for 111
countries. We employ 4 of these dimensions, which are correlated to
the costumers- buying behavior, in order to detect users- preferences
for items. In addition, several advantages of this method
demonstrated for data sparseness and cold-start users, which are
important challenges in collaborative filtering. We present
experiments using a real dataset, Book Crossing Dataset.
Experimental results shows that the proposed algorithm provide
significant advantages in terms of improving recommendation
quality.
Abstract: The implementation of Super-Ultra Low Emission
Vehicle standards requires more efficient exhaust gas purification. To
increase the efficiency of exhaust gas purification, an the adsorbent
capable of holding hydrocarbons up to 250-300 ОС should be
developed. The possibility to design such adsorbents by modification
of zeolites of mordenite type, ZSM-5 and NaY, using different
metals cations has been studied.
It has been shown that introducing Cr, Cs, Zn, Ni, Co, Li, Mn in
zeolites results in modification of the toluene TPD and toluene
sorption capacity.
5%LiZSM-5 zeolite exhibits the most attractive TPD curve, with
toluene desorption temperature ranging from 250 to 350ОС. The
sorption capacity of 5%Li-ZSM-5 is 0.4 mmol/g. NaY zeolite has the
highest sorption capacity, up to 2 mmol/g, and holds toluene up to
350ОС, but at 120ОС toluene desorption starts, which is not desirable,
since the adsorbent of cold start hydrocarbons should retain them
until 250-300ОС. Therefore 5%LiZSM-5 zeolite was found to be the
most promising to control the cold-start hydrocarbon emissions
among the samples studied.